Update medical_rag.py

medical_rag.py

@@ -104,55 +104,108 @@ def create_medical_prompt(question, passage):
 @app.post("/api/chat")
 async def chat_endpoint(chat_query: ChatQuery):
     try:
+        print("\n=== STARTING CHAT REQUEST PROCESSING ===")
+        print(f"Initial query: {chat_query.query} (language_code: {chat_query.language_code})")
+        
+        # Step 1: Handle translation if needed
         query_text = chat_query.query
-        language_code = chat_query.language_code        
+        language_code = chat_query.language_code
         if language_code == 0:
+            print("Translating from Arabic to English...")
             query_text = translate_text(query_text, 'ar_to_en')
+            print(f"Translated query: {query_text}")
         
-        # Generate embeddings and retrieve relevant documents
+        # Step 2: Generate embeddings
+        print("\nGenerating query embeddings...")
         query_embedding = embed_query_text(query_text)
-        n_results = 5
+        print(f"Embedding generated. Shape: {query_embedding.shape}")
+        
+        # Step 3: Load embeddings and query them
+        print("\nLoading document embeddings...")
         embeddings_data = load_embeddings()
-        folder_path = 'downloaded_articles/downloaded_articles'
+        if not embeddings_data:
+            raise HTTPException(status_code=500, detail="Failed to load embeddings data")
+        print(f"Loaded embeddings for {len(embeddings_data)} documents")
+        
+        print("\nQuerying embeddings...")
+        n_results = 5
         initial_results = query_embeddings(query_embedding, embeddings_data, n_results)
+        print(f"Initial results: {initial_results}")
+        
         document_ids = [doc_id for doc_id, _ in initial_results]
+        print(f"Document IDs to retrieve: {document_ids}")
+        
+        # Step 4: Retrieve document texts
+        print("\nRetrieving document texts...")
+        folder_path = 'downloaded_articles/downloaded_articles'
         document_texts = retrieve_document_texts(document_ids, folder_path)
+        print(f"Retrieved {len(document_texts)} documents")
+        
+        # Step 5: Rerank documents
+        print("\nReranking documents...")
+        if 'cross_encoder' not in models:
+            raise HTTPException(status_code=500, detail="Cross-encoder model not loaded")
         
-        # Rerank documents with cross-encoder
         cross_encoder = models['cross_encoder']
         scores = cross_encoder.predict([(query_text, doc) for doc in document_texts])
         scored_documents = list(zip(scores, document_ids, document_texts))
         scored_documents.sort(key=lambda x: x[0], reverse=True)
+        print("Top 3 reranked documents:")
+        for i, (score, doc_id, _) in enumerate(scored_documents[:3]):
+            print(f"{i+1}. Doc {doc_id} (score: {score:.4f})")
         
-        # Extract relevant portions from documents using medical-specific function
+        # Step 6: Extract relevant portions
+        print("\nExtracting relevant portions...")
         relevant_portions = extract_relevant_portions(document_texts, query_text)
+        print(f"Found relevant portions in {len(relevant_portions)} documents")
+        
         flattened_relevant_portions = []
         for doc_id, portions in relevant_portions.items():
             flattened_relevant_portions.extend(portions)
         
         combined_parts = " ".join(flattened_relevant_portions)
+        print(f"Combined relevant text length: {len(combined_parts)} characters")
+        
+        # Step 7: Extract and enhance with entities
+        print("\nExtracting entities...")
         entities = extract_entities(query_text)
+        print(f"Found entities: {entities}")
+        
         passage = enhance_passage_with_entities(combined_parts, entities)
+        print(f"Enhanced passage length: {len(passage)} characters")
         
-        # Create medical-specific prompt and get completion from DeepSeek
+        # Step 8: Generate response
+        print("\nCreating prompt...")
         prompt = create_medical_prompt(query_text, passage)
+        print(f"Prompt length: {len(prompt)} characters")
+        
+        print("\nGetting completion from DeepSeek...")
         answer = get_completion(prompt)
+        print(f"Raw answer received: {answer[:200]}...")  # Print first 200 chars
         
+        # Step 9: Final processing
         final_answer = answer.strip()
         if language_code == 0:
+            print("\nTranslating answer to Arabic...")
             final_answer = translate_text(final_answer, 'en_to_ar')
+            print(f"Translated answer: {final_answer[:200]}...")
             
         if not final_answer:
             final_answer = "Sorry, I can't help with that."
-            
+            print("Warning: Empty answer received")
+        
+        print("\n=== REQUEST PROCESSING COMPLETE ===")
         return {
             "response": f"I hope this answers your question: {final_answer}",
             "success": True
         }
         
     except HTTPException as e:
+        print(f"\n!!! HTTPException: {e.detail}")
         raise e
     except Exception as e:
+        print(f"\n!!! Unexpected error: {str(e)}")
+        print(f"Error type: {type(e).__name__}")
         raise HTTPException(status_code=500, detail=str(e))
 
 # Initialize medical models when this module is imported